High-dimensional inference
Statistical methods with principled guarantees for modern, structured, and high-dimensional data.
STATISTICS · MACHINE LEARNING
Seungsu Han
Incoming Ph.D. Student in Operations Research and Financial Engineering at Princeton University
I study statistically principled machine learning, with a focus on high-dimensional inference and uncertainty quantification.
RESEARCH THEMES
Reliable inference in complex models
Uncertainty quantification
Flexible variational approximations that preserve multimodality, dependence, and heavy-tail behavior.
Robust learning
Methods that remain reliable under misspecification, distribution shift, and non-Gaussian structure.
SELECTED WORK
Publications
2025
Stick-Breaking Mixture Normalizing Flows with Component-Wise Tail Adaptation for Variational Inference
S. Han, J. Hwang, and W. Chang
arXiv:2510.07965 · Under review
2025
Optimal Estimation of Linear Non-Gaussian Structural Equation Models
S. Oh, S. Han, and G. Park
Proceedings of AISTATS 2025
RECENT HIGHLIGHTS
News
Joining Princeton University as a Ph.D. student in Operations Research and Financial Engineering.
Presented StiCTAF at the Joint International Seminar with Kyushu University in Fukuoka, Japan.
Presented our work on optimal estimation of LiNGAMs at AISTATS 2025 in Phuket, Thailand.